Asymmetry in the complexity of same and different representations

Visual reasoning in object-centric deep neural networks: A comparative cognition approach

Achieving visual reasoning is a long-term goal of artificial intelligence. In the last decade, several studies have applied deep neural networks (DNNs) to the task of learning visual relations from images, with modest results in terms of generalization of the relations learned. However, in recent years, object-centric representation learning has been put forward as a way to achieve visual reasoning within the deep learning framework. Object-centric models attempt to model input scenes as compositions of objects and relations between them. To this end, these models use several kinds of attention mechanisms to segregate the individual objects in a scene from the background and from other objects. In this work we tested relation learning and generalization in several object-centric models, as well as a ResNet-50 baseline. In contrast to previous research, which has focused heavily in the same-different task in order to asses relational reasoning in DNNs, we use a set of tasks - with varying degrees of complexity - derived from the comparative cognition literature. Our results show that object-centric models are able to segregate the different objects in a scene, even in many out-of-distribution cases. In our simpler tasks, this improves their capacity to learn and generalize visual relations in comparison to the ResNet-50 baseline. However, object-centric models still struggle in our more difficult tasks and conditions. We conclude that abstract visual reasoning remains an open challenge for DNNs, including object-centric models.

Stimulus-dependent emergence of understanding the 'same-different' concept in budgerigars

The ability to understand relational concepts, such as 'same' and 'different', is a critical feature of human cognition. To what extent non-human animals can acquire such concepts and which factors influence their learning are still unclear. We examined the acquisition and the breadth of understanding the 'same-different' concept in budgerigars (Melopsittacus undulatus). Budgerigars trained to discriminate stimulus pairs in which two identical figures were either the same or different size (Experiment 1) successfully generalized the discrimination to novel stimuli belonging to various categories (size, colour, shape, geometric type and number of dots). The results of Experiment 1 thus demonstrate that budgerigars can perceive and generalize the same-different concept across dimensions after training with a limited set of stimuli differing along a single dimension. In contrast, while most budgerigars trained to discriminate two pairs of discs that were either the same or different in colour (Experiment 2) could generalize the discrimination to novel stimuli within the training category (colour), only few generalized the discrimination to another category suggesting a generalization based on perceptual similarity. The results thus show that whether budgerigars generalize a relationship by conceptual or perceptual similarity depends on the nature of the training stimuli.

Cognitive complexity explains processing asymmetry in judgments of similarity versus difference

Human judgments of similarity and difference are sometimes asymmetrical, with the former being more sensitive than the latter to relational overlap, but the theoretical basis for this asymmetry remains unclear. We test an explanation based on the type of information used to make these judgments (relations versus features) and the comparison process itself (similarity versus difference). We propose that asymmetries arise from two aspects of cognitive complexity that impact judgments of similarity and difference: processing relations between entities is more cognitively demanding than processing features of individual entities, and comparisons assessing difference are more cognitively complex than those assessing similarity. In Experiment 1 we tested this hypothesis for both verbal comparisons between word pairs, and visual comparisons between sets of geometric shapes. Participants were asked to select one of two options that was either more similar to or more different from a standard. On unambiguous trials, one option was unambiguously more similar to the standard; on ambiguous trials, one option was more featurally similar to the standard, whereas the other was more relationally similar. Given the higher cognitive complexity of processing relations and of assessing difference, we predicted that detecting relational difference would be particularly demanding. We found that participants (1) had more difficulty detecting relational difference than they did relational similarity on unambiguous trials, and (2) tended to emphasize relational information more when judging similarity than when judging difference on ambiguous trials. The latter finding was replicated using more complex story stimuli (Experiment 2). We showed that this pattern can be captured by a computational model of comparison that weights relational information more heavily for similarity than for difference judgments.

Ask me why, don't tell me why: Asking children for explanations facilitates relational thinking

Identifying abstract relations is essential for commonsense reasoning. Research suggests that even young children can infer relations such as "same" and "different," but often fail to apply these concepts. Might the process of explaining facilitate the recognition and application of relational concepts? Based on prior work suggesting that explanation can be a powerful tool to promote abstract reasoning, we predicted that children would be more likely to discover and use an abstract relational rule when they were prompted to explain observations instantiating that rule, compared to when they received demonstration alone. Five- and 6-year-olds were given a modified Relational Match to Sample (RMTS) task, with repeated demonstrations of relational (same) matches by an adult. Half of the children were prompted to explain these matches; the other half reported the match they observed. Children who were prompted to explain showed immediate, stable success, while those only asked to report the outcome of the pedagogical demonstration did not. Findings provide evidence that explanation facilitates early abstraction over and above demonstration alone.

Abstract representations of small sets in newborns

From the very first days of life, newborns are not tied to represent narrow, modality- and object-specific aspects of their environment. Rather, they sometimes react to abstract properties shared by stimuli of very different nature, such as approximate numerosity or magnitude. As of now, however, there is no evidence that newborns possess abstract representations that apply to small sets: in particular, while newborns can match large approximate numerosities across senses, this ability does not extend to small numerosities. In two experiments, we presented newborn infants (N = 64, age 17 to 98 h) with patterned sets AB or ABB simultaneously in the auditory and visual modalities. Auditory patterns were presented as periodic sequences of sounds (AB: triangle-drum-triangle-drum-triangle-drum …; ABB: triangle-drum-drum-triangle-drum-drum-triangle-drum-drum …), and visual patterns as arrays of 2 or 3 shapes (AB: circle-diamond; ABB: circle-diamond-diamond). In both experiments, we found that participants reacted and looked longer when the patterns matched across the auditory and visual modalities – provided that the first stimulus they received was congruent. These findings uncover the existence of yet another type of abstract representations at birth, applying to small sets. As such, they bolster the hypothesis that newborns are endowed with the capacity to represent their environment in broad strokes, in terms of its most abstract properties. This capacity for abstraction could later serve as a scaffold for infants to learn about the particular entities surrounding them.

•

Newborns were presented with auditory and visual patterns (AB vs. ABB).

•

Participants reacted when the patterns presented were congruent across modalities.

•

Newborns possess abstract representations applying to small sets.

•

These representations may encode numerosity and/or repetitions.

Representations of Abstract Relations in Infancy

relations are considered the pinnacle of human cognition, allowing for analogical and logical reasoning, and possibly setting humans apart from other animal species. Recent experimental evidence showed that infants are capable of representing the abstract relations same and different, prompting the question of the format of such representations. In a propositional language of thought, abstract relations would be represented in the form of discrete symbols. Is this format available to pre-lexical infants? We report six experiments (N = 192) relying on pupillometry and investigating how preverbal 10- to 12-month-old infants represent the relation same. We found that infants' ability to represent the relation same is impacted by the number of individual entities taking part in the relation. Infants could represent that four syllables were the same and generalized that relation to novel sequences (Experiments 1 and 4). However, they failed to generalize the relation same when it involved 5 or 6 syllables (Experiments 2-3), showing that infants' representation of the relation same is constrained by the limits of working memory capacity. Infants also failed to form a representation equivalent to all the same, which could apply to a varying number of same syllables (Experiments 5-6). These results highlight important discontinuities along cognitive development. Contrary to adults, preverbal infants lack a discrete symbol for the relation same, and rather build a representation of the relation by assembling symbols for individual entities.

It's not all the same to pigeons: Representations of difference may be shared across species

Pigeons readily learn and transfer same-different discriminations in a variety of experimental paradigms. However, strategically designed probe tests suggest that they might only represent sameness. Here, we provide the first direct evidence that pigeons also represent difference. We first trained pigeons on a conditional same-different discrimination; then, on probe trials, we replaced either the same-item pair or the different-item pair with a familiar, but ambiguous stimulus. On different-cued probe trials, pigeons' choices were controlled by sameness: they reliably rejected the same-item pair, but they did not reliably select the different-item pair. Conversely, on same-cued probe trials, pigeons' choices were controlled by difference: they reliably rejected the different-item pair, but they did not reliably select the same-item pair. Together, these findings demonstrate that pigeons can represent both sameness and difference, providing an important clue to elucidating the evolutionary origins of same-different conceptualization.

Negative mental representations in infancy

How do infants' thoughts compare to the thoughts adults express with language? In particular, can infants entertain negative representations, such as not red or not here? In four experiments, we used pupillometry to ask whether negative representations are possible without an external language. Eleven-month-olds were tested on their ability to detect and represent the abstract structure of sequences of syllables, defined by the relations identity and/or negation: AAAA (four identical syllables; Experiment 1), AAA¬A (three times the syllable A and one final syllable that is not A; Experiment 2), AA(A)(A)¬A (two-to-four times the syllable A and one final syllable that is not A; Experiment 3). Representing the structures in Experiments 2-3 requires a form of negation. Results suggest that infants are able to compute both identity and negation. More generally, these results lend credit to the hypothesis that the infant mind is equipped with rudimentary logical operators before language takes off.